linux-toradex.git/net/core, branch v5.3-rc8

net: fix skb use after free in netpoll

2019-08-28T03:52:02+00:00

After commit baeababb5b85d5c4e6c917efe2a1504179438d3b
("tun: return NET_XMIT_DROP for dropped packets"),
when tun_net_xmit drop packets, it will free skb and return NET_XMIT_DROP,
netpoll_send_skb_on_dev will run into following use after free cases:
1. retry netpoll_start_xmit with freed skb;
2. queue freed skb in npinfo->txq.
queue_process will also run into use after free case.

hit netpoll_send_skb_on_dev first case with following kernel log:

[  117.864773] kernel BUG at mm/slub.c:306!
[  117.864773] invalid opcode: 0000 [#1] SMP PTI
[  117.864774] CPU: 3 PID: 2627 Comm: loop_printmsg Kdump: loaded Tainted: P           OE     5.3.0-050300rc5-generic #201908182231
[  117.864775] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014
[  117.864775] RIP: 0010:kmem_cache_free+0x28d/0x2b0
[  117.864781] Call Trace:
[  117.864781]  ? tun_net_xmit+0x21c/0x460
[  117.864781]  kfree_skbmem+0x4e/0x60
[  117.864782]  kfree_skb+0x3a/0xa0
[  117.864782]  tun_net_xmit+0x21c/0x460
[  117.864782]  netpoll_start_xmit+0x11d/0x1b0
[  117.864788]  netpoll_send_skb_on_dev+0x1b8/0x200
[  117.864789]  __br_forward+0x1b9/0x1e0 [bridge]
[  117.864789]  ? skb_clone+0x53/0xd0
[  117.864790]  ? __skb_clone+0x2e/0x120
[  117.864790]  deliver_clone+0x37/0x50 [bridge]
[  117.864790]  maybe_deliver+0x89/0xc0 [bridge]
[  117.864791]  br_flood+0x6c/0x130 [bridge]
[  117.864791]  br_dev_xmit+0x315/0x3c0 [bridge]
[  117.864792]  netpoll_start_xmit+0x11d/0x1b0
[  117.864792]  netpoll_send_skb_on_dev+0x1b8/0x200
[  117.864792]  netpoll_send_udp+0x2c6/0x3e8
[  117.864793]  write_msg+0xd9/0xf0 [netconsole]
[  117.864793]  console_unlock+0x386/0x4e0
[  117.864793]  vprintk_emit+0x17e/0x280
[  117.864794]  vprintk_default+0x29/0x50
[  117.864794]  vprintk_func+0x4c/0xbc
[  117.864794]  printk+0x58/0x6f
[  117.864795]  loop_fun+0x24/0x41 [printmsg_loop]
[  117.864795]  kthread+0x104/0x140
[  117.864795]  ? 0xffffffffc05b1000
[  117.864796]  ? kthread_park+0x80/0x80
[  117.864796]  ret_from_fork+0x35/0x40

Signed-off-by: Feng Sun 
Signed-off-by: Xiaojun Zhao 
Signed-off-by: David S. Miller

sock: fix potential memory leak in proto_register()

2019-08-24T23:33:14+00:00

If protocols registered exceeded PROTO_INUSE_NR, prot will be
added to proto_list, but no available bit left for prot in
proto_inuse_idx.

Changes since v2:
* Propagate the error code properly

Signed-off-by: zhanglin 
Signed-off-by: David S. Miller

Merge git://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf

2019-08-24T00:34:11+00:00

Daniel Borkmann says:

====================
pull-request: bpf 2019-08-24

The following pull-request contains BPF updates for your *net* tree.

The main changes are:

1) Fix verifier precision tracking with BPF-to-BPF calls, from Alexei.

2) Fix a use-after-free in prog symbol exposure, from Daniel.

3) Several s390x JIT fixes plus BE related fixes in BPF kselftests, from Ilya.

4) Fix memory leak by unpinning XDP umem pages in error path, from Ivan.

5) Fix a potential use-after-free on flow dissector detach, from Jakub.

6) Fix bpftool to close prog fd after showing metadata, from Quentin.

7) BPF kselftest config and TEST_PROGS_EXTENDED fixes, from Anders.
====================

Signed-off-by: David S. Miller

bpf: allow narrow loads of some sk_reuseport_md fields with offset > 0

2019-08-23T23:25:41+00:00

test_select_reuseport fails on s390 due to verifier rejecting
test_select_reuseport_kern.o with the following message:

	; data_check.eth_protocol = reuse_md->eth_protocol;
	18: (69) r1 = *(u16 *)(r6 +22)
	invalid bpf_context access off=22 size=2

This is because on big-endian machines casts from __u32 to __u16 are
generated by referencing the respective variable as __u16 with an offset
of 2 (as opposed to 0 on little-endian machines).

The verifier already has all the infrastructure in place to allow such
accesses, it's just that they are not explicitly enabled for
eth_protocol field. Enable them for eth_protocol field by using
bpf_ctx_range instead of offsetof.

Ditto for ip_protocol, bind_inany and len, since they already allow
narrowing, and the same problem can arise when working with them.

Fixes: 2dbb9b9e6df6 ("bpf: Introduce BPF_PROG_TYPE_SK_REUSEPORT")
Signed-off-by: Ilya Leoshkevich 
Signed-off-by: Daniel Borkmann

flow_dissector: Fix potential use-after-free on BPF_PROG_DETACH

2019-08-23T23:15:34+00:00

Call to bpf_prog_put(), with help of call_rcu(), queues an RCU-callback to
free the program once a grace period has elapsed. The callback can run
together with new RCU readers that started after the last grace period.
New RCU readers can potentially see the "old" to-be-freed or already-freed
pointer to the program object before the RCU update-side NULLs it.

Reorder the operations so that the RCU update-side resets the protected
pointer before the end of the grace period after which the program will be
freed.

Fixes: d58e468b1112 ("flow_dissector: implements flow dissector BPF hook")
Reported-by: Lorenz Bauer 
Signed-off-by: Jakub Sitnicki 
Acked-by: Petar Penkov 
Signed-off-by: Daniel Borkmann

tcp: make sure EPOLLOUT wont be missed

2019-08-19T20:07:43+00:00

As Jason Baron explained in commit 790ba4566c1a ("tcp: set SOCK_NOSPACE
under memory pressure"), it is crucial we properly set SOCK_NOSPACE
when needed.

However, Jason patch had a bug, because the 'nonblocking' status
as far as sk_stream_wait_memory() is concerned is governed
by MSG_DONTWAIT flag passed at sendmsg() time :

    long timeo = sock_sndtimeo(sk, flags & MSG_DONTWAIT);

So it is very possible that tcp sendmsg() calls sk_stream_wait_memory(),
and that sk_stream_wait_memory() returns -EAGAIN with SOCK_NOSPACE
cleared, if sk->sk_sndtimeo has been set to a small (but not zero)
value.

This patch removes the 'noblock' variable since we must always
set SOCK_NOSPACE if -EAGAIN is returned.

It also renames the do_nonblock label since we might reach this
code path even if we were in blocking mode.

Fixes: 790ba4566c1a ("tcp: set SOCK_NOSPACE under memory pressure")
Signed-off-by: Eric Dumazet 
Cc: Jason Baron 
Reported-by: Vladimir Rutsky  
Acked-by: Soheil Hassas Yeganeh 
Acked-by: Neal Cardwell 
Acked-by: Jason Baron 
Signed-off-by: David S. Miller

sock: make cookie generation global instead of per netns

2019-08-09T20:14:46+00:00

Generating and retrieving socket cookies are a useful feature that is
exposed to BPF for various program types through bpf_get_socket_cookie()
helper.

The fact that the cookie counter is per netns is quite a limitation
for BPF in practice in particular for programs in host namespace that
use socket cookies as part of a map lookup key since they will be
causing socket cookie collisions e.g. when attached to BPF cgroup hooks
or cls_bpf on tc egress in host namespace handling container traffic
from veth or ipvlan devices with peer in different netns. Change the
counter to be global instead.

Socket cookie consumers must assume the value as opqaue in any case.
Not every socket must have a cookie generated and knowledge of the
counter value itself does not provide much value either way hence
conversion to global is fine.

Signed-off-by: Daniel Borkmann 
Cc: Eric Dumazet 
Cc: Alexei Starovoitov 
Cc: Willem de Bruijn 
Cc: Martynas Pumputis 
Signed-off-by: David S. Miller

net/tls: prevent skb_orphan() from leaking TLS plain text with offload

2019-08-09T05:39:35+00:00

sk_validate_xmit_skb() and drivers depend on the sk member of
struct sk_buff to identify segments requiring encryption.
Any operation which removes or does not preserve the original TLS
socket such as skb_orphan() or skb_clone() will cause clear text
leaks.

Make the TCP socket underlying an offloaded TLS connection
mark all skbs as decrypted, if TLS TX is in offload mode.
Then in sk_validate_xmit_skb() catch skbs which have no socket
(or a socket with no validation) and decrypted flag set.

Note that CONFIG_SOCK_VALIDATE_XMIT, CONFIG_TLS_DEVICE and
sk->sk_validate_xmit_skb are slightly interchangeable right now,
they all imply TLS offload. The new checks are guarded by
CONFIG_TLS_DEVICE because that's the option guarding the
sk_buff->decrypted member.

Second, smaller issue with orphaning is that it breaks
the guarantee that packets will be delivered to device
queues in-order. All TLS offload drivers depend on that
scheduling property. This means skb_orphan_partial()'s
trick of preserving partial socket references will cause
issues in the drivers. We need a full orphan, and as a
result netem delay/throttling will cause all TLS offload
skbs to be dropped.

Reusing the sk_buff->decrypted flag also protects from
leaking clear text when incoming, decrypted skb is redirected
(e.g. by TC).

See commit 0608c69c9a80 ("bpf: sk_msg, sock{map|hash} redirect
through ULP") for justification why the internal flag is safe.
The only location which could leak the flag in is tcp_bpf_sendmsg(),
which is taken care of by clearing the previously unused bit.

v2:
 - remove superfluous decrypted mark copy (Willem);
 - remove the stale doc entry (Boris);
 - rely entirely on EOR marking to prevent coalescing (Boris);
 - use an internal sendpages flag instead of marking the socket
   (Boris).
v3 (Willem):
 - reorganize the can_skb_orphan_partial() condition;
 - fix the flag leak-in through tcp_bpf_sendmsg.

Signed-off-by: Jakub Kicinski 
Acked-by: Willem de Bruijn 
Reviewed-by: Boris Pismenny 
Signed-off-by: David S. Miller

net: fix bpf_xdp_adjust_head regression for generic-XDP

2019-08-05T18:17:40+00:00

When generic-XDP was moved to a later processing step by commit
458bf2f224f0 ("net: core: support XDP generic on stacked devices.")
a regression was introduced when using bpf_xdp_adjust_head.

The issue is that after this commit the skb->network_header is now
changed prior to calling generic XDP and not after. Thus, if the header
is changed by XDP (via bpf_xdp_adjust_head), then skb->network_header
also need to be updated again.  Fix by calling skb_reset_network_header().

Fixes: 458bf2f224f0 ("net: core: support XDP generic on stacked devices.")
Reported-by: Brandon Cazander 
Signed-off-by: Jesper Dangaard Brouer 
Signed-off-by: David S. Miller

net: fix ifindex collision during namespace removal

2019-07-29T18:07:04+00:00

Commit aca51397d014 ("netns: Fix arbitrary net_device-s corruptions
on net_ns stop.") introduced a possibility to hit a BUG in case device
is returning back to init_net and two following conditions are met:
1) dev->ifindex value is used in a name of another "dev%d"
   device in init_net.
2) dev->name is used by another device in init_net.

Under real life circumstances this is hard to get. Therefore this has
been present happily for over 10 years. To reproduce:

$ ip a
1: lo:  mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
       valid_lft forever preferred_lft forever
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
2: dummy0:  mtu 1500 qdisc noop state DOWN group default qlen 1000
    link/ether 86:89:3f:86:61:29 brd ff:ff:ff:ff:ff:ff
3: enp0s2:  mtu 1500 qdisc noop state DOWN group default qlen 1000
    link/ether 52:54:00:12:34:56 brd ff:ff:ff:ff:ff:ff
$ ip netns add ns1
$ ip -n ns1 link add dummy1ns1 type dummy
$ ip -n ns1 link add dummy2ns1 type dummy
$ ip link set enp0s2 netns ns1
$ ip -n ns1 link set enp0s2 name dummy0
[  100.858894] virtio_net virtio0 dummy0: renamed from enp0s2
$ ip link add dev4 type dummy
$ ip -n ns1 a
1: lo:  mtu 65536 qdisc noop state DOWN group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
2: dummy1ns1:  mtu 1500 qdisc noop state DOWN group default qlen 1000
    link/ether 16:63:4c:38:3e:ff brd ff:ff:ff:ff:ff:ff
3: dummy2ns1:  mtu 1500 qdisc noop state DOWN group default qlen 1000
    link/ether aa:9e:86:dd:6b:5d brd ff:ff:ff:ff:ff:ff
4: dummy0:  mtu 1500 qdisc noop state DOWN group default qlen 1000
    link/ether 52:54:00:12:34:56 brd ff:ff:ff:ff:ff:ff
$ ip a
1: lo:  mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
       valid_lft forever preferred_lft forever
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
2: dummy0:  mtu 1500 qdisc noop state DOWN group default qlen 1000
    link/ether 86:89:3f:86:61:29 brd ff:ff:ff:ff:ff:ff
4: dev4:  mtu 1500 qdisc noop state DOWN group default qlen 1000
    link/ether 5a:e1:4a:b6:ec:f8 brd ff:ff:ff:ff:ff:ff
$ ip netns del ns1
[  158.717795] default_device_exit: failed to move dummy0 to init_net: -17
[  158.719316] ------------[ cut here ]------------
[  158.720591] kernel BUG at net/core/dev.c:9824!
[  158.722260] invalid opcode: 0000 [#1] SMP KASAN PTI
[  158.723728] CPU: 0 PID: 56 Comm: kworker/u2:1 Not tainted 5.3.0-rc1+ #18
[  158.725422] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.12.0-2.fc30 04/01/2014
[  158.727508] Workqueue: netns cleanup_net
[  158.728915] RIP: 0010:default_device_exit.cold+0x1d/0x1f
[  158.730683] Code: 84 e8 18 c9 3e fe 0f 0b e9 70 90 ff ff e8 36 e4 52 fe 89 d9 4c 89 e2 48 c7 c6 80 d6 25 84 48 c7 c7 20 c0 25 84 e8 f4 c8 3e
[  158.736854] RSP: 0018:ffff8880347e7b90 EFLAGS: 00010282
[  158.738752] RAX: 000000000000003b RBX: 00000000ffffffef RCX: 0000000000000000
[  158.741369] RDX: 0000000000000000 RSI: ffffffff8128013d RDI: ffffed10068fcf64
[  158.743418] RBP: ffff888033550170 R08: 000000000000003b R09: fffffbfff0b94b9c
[  158.745626] R10: fffffbfff0b94b9b R11: ffffffff85ca5cdf R12: ffff888032f28000
[  158.748405] R13: dffffc0000000000 R14: ffff8880335501b8 R15: 1ffff110068fcf72
[  158.750638] FS:  0000000000000000(0000) GS:ffff888036000000(0000) knlGS:0000000000000000
[  158.752944] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  158.755245] CR2: 00007fe8b45d21d0 CR3: 00000000340b4005 CR4: 0000000000360ef0
[  158.757654] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[  158.760012] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
[  158.762758] Call Trace:
[  158.763882]  ? dev_change_net_namespace+0xbb0/0xbb0
[  158.766148]  ? devlink_nl_cmd_set_doit+0x520/0x520
[  158.768034]  ? dev_change_net_namespace+0xbb0/0xbb0
[  158.769870]  ops_exit_list.isra.0+0xa8/0x150
[  158.771544]  cleanup_net+0x446/0x8f0
[  158.772945]  ? unregister_pernet_operations+0x4a0/0x4a0
[  158.775294]  process_one_work+0xa1a/0x1740
[  158.776896]  ? pwq_dec_nr_in_flight+0x310/0x310
[  158.779143]  ? do_raw_spin_lock+0x11b/0x280
[  158.780848]  worker_thread+0x9e/0x1060
[  158.782500]  ? process_one_work+0x1740/0x1740
[  158.784454]  kthread+0x31b/0x420
[  158.786082]  ? __kthread_create_on_node+0x3f0/0x3f0
[  158.788286]  ret_from_fork+0x3a/0x50
[  158.789871] ---[ end trace defd6c657c71f936 ]---
[  158.792273] RIP: 0010:default_device_exit.cold+0x1d/0x1f
[  158.795478] Code: 84 e8 18 c9 3e fe 0f 0b e9 70 90 ff ff e8 36 e4 52 fe 89 d9 4c 89 e2 48 c7 c6 80 d6 25 84 48 c7 c7 20 c0 25 84 e8 f4 c8 3e
[  158.804854] RSP: 0018:ffff8880347e7b90 EFLAGS: 00010282
[  158.807865] RAX: 000000000000003b RBX: 00000000ffffffef RCX: 0000000000000000
[  158.811794] RDX: 0000000000000000 RSI: ffffffff8128013d RDI: ffffed10068fcf64
[  158.816652] RBP: ffff888033550170 R08: 000000000000003b R09: fffffbfff0b94b9c
[  158.820930] R10: fffffbfff0b94b9b R11: ffffffff85ca5cdf R12: ffff888032f28000
[  158.825113] R13: dffffc0000000000 R14: ffff8880335501b8 R15: 1ffff110068fcf72
[  158.829899] FS:  0000000000000000(0000) GS:ffff888036000000(0000) knlGS:0000000000000000
[  158.834923] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
[  158.838164] CR2: 00007fe8b45d21d0 CR3: 00000000340b4005 CR4: 0000000000360ef0
[  158.841917] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[  158.845149] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400

Fix this by checking if a device with the same name exists in init_net
and fallback to original code - dev%d to allocate name - in case it does.

This was found using syzkaller.

Fixes: aca51397d014 ("netns: Fix arbitrary net_device-s corruptions on net_ns stop.")
Signed-off-by: Jiri Pirko 
Signed-off-by: David S. Miller